On Wednesday, 30 September 2015 at 19:24:05 UTC, Jeremy DeHaan
wrote:
On Wednesday, 30 September 2015 at 17:51:50 UTC, Jan Johansson
wrote:
[...]
Like Adam said, the real difference between a .d and a .di file
is that the .di file has all the guts removed and is just the
declarations.
If using a .di file is really what you want, you could try
something like this?
test.d:
module test;
interface IMyTest {
void Write(string message);
}
IMyTest createInstance() {
class MyTest : IMyTest {
void Write(string message) {
import std.stdio;
writeln(message);
}
}
return new MyTest;
}
---------------
test.di:
module test;
interface IMyTest {
void Write(string message);
}
IMyTest createInstance();
---------------
main.d:
import test;
void main() {
auto p = createInstance();
p.Write("Hello, World!");
}
--------------
and then
dmd test.d -lib -oftest
and
dmd main.d test.di test.a
Also like Adam said, dmd can create these .di files for you so
you don't have to!
(This is untested, but should work/be close to working)
Thanks Jeremy,
Do you spot a weakness in your proposed code snip? The
declaration for interface is done in two separate files, both
test.d and test.di. Scattered declarations has never been a good
idea. I know that I can ask the DMD to do declaration files for
me, but the use case for that is to speed up building of
executable.
But is it that the separation of declaration and implementation
was never the intention in the design of D?
//Jan
